Remote Sensing Estimation of Chlorophyll-A in Case-II Waters of Coastal Areas: Three-Band Model Versus Genetic Algorithm–Artificial Neural Networks Model
Jinyue Chen,Shuisen Chen,Rao Fu,Chongyang Wang,Dan Li,Yongshi Peng,Li Wang,Hao Jiang,Qiong Zheng
DOI: https://doi.org/10.1109/jstars.2021.3066697
IF: 4.715
2021-01-01
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Abstract:Chlorophyll-a (Chl-a), an important indicator of phytoplankton biomass and eutrophication, is sensitive to water constitutes and optical characteristics. An integrated machine learning method of genetic algorithm and artificial neural networks (GA–ANN) was developed to retrieve the concentration of Chl-a. In situ spectra and simultaneous water quality parameters of 107 samples from two reservoirs (Res) and coastal waters (CW) were used to calibrate GA–ANN and three-band models (TBM) for comparison of Chl-a estimation. Both GA–ANN and TBM methods perform well for the joint dataset (WGD) of Res and CW with the R<sup>2</sup> exceeding 0.90, and the root mean square error (RMSE) of corresponding validation (N = 35) are 4.40 and 5.23 μg/L, respectively. Similarly, for independent dataset of Res (N = 45), GA–ANN and TBM methods show robust performance: the R<sup>2</sup> values are 0.87 and 0.80, respectively; and the corresponding RMSE values are 7.79 and 7.73 μg/L, respectively. For CW dataset (N = 62), the R<sup>2</sup> values of two methods are 0.81 and 0.62, respectively; and the corresponding RMSE values are 0.79 and 1.32 μg/L, respectively. When the GA–ANN and TBM models were applied to retrieve Chl-a concentration from the calibrated Sentinel 2 MSI reflectance data in two Res on October 20, 2019, however, the validated results of MSI-derived Chl-a concentrations using quasi-synchronous in situ data (N = 36) indicated that the GA–ANN model out-erforms TBM with higher R<sup>2</sup> value (0.91 vs. 0.26) and smaller RMSE (4.41 vs. 13.85 μg/L) and mean absolute errors (3.40 vs. 11.87 μg/L) values. Although TBM has obvious overestimation of Chl-a concentration when applied to remote sensing image, we still thought that both GA–ANN and TBM are useful methods for Chl-a estimation in case-II waters, and GA–ANN performs marginally better with less deviation to measured Chl-a for multispectral remote sensing data. The ratio of TSS to Chl-a, experimental measurements, abundance of sampling points, and Chl-a concentration range are several important factors affecting the accuracy and robustness of GA–ANN and TBM methods.
imaging science & photographic technology,remote sensing,engineering, electrical & electronic,geography, physical